DAFX: Digital Audio Effects, Second Edition

9.5 Conclusions

From existing simple effects such as the compressor and auto-tune, we presented a general framework for adaptive DAFX using sound features as control parameters. Such adaptive control can rely on any type of sound features that are extracted from various representations of sound: samples, STFT, spectrum, source and filter. Indeed, previously described techniques such as the phase vocoder, source-filter models, and spectral models do allow for both the representation of the sound and for the extraction of further global features such as pitch or fundamental frequency, which can be estimated by the cepstrum method or auto-correlation techniques applied to the input directly or the extracted source signal. Further global features such as amplitude envelope, spectral centroid, and auto-correlation features (voiced/unvoiced detection) have been introduced, which can be estimated by simple time-domain or by advanced time-frequency techniques.

1 Recent literature tends to use ‘sound descriptors’ instead of ‘sound features’; both describe the same reality, that is to say parameters that describe the sound signal, and/or some of its statistical, acoustical or perceptual properties.

2 http://psysound.wikidot.com/

3 M is the number of features we use, usually between 1 and 5; N is the number of effect control parameters, usually between 1 and 20.

4 There is no analytical solution, so an iterative scheme is necessary.

5 In order to avoid rapid pitch-shifting modifications at ...